Machine Learning Methods to Analyze Healthcare Survey Data Sets

Abstract

The utilization of machine learning in the healthcare sector is transforming the examination of patient satisfaction data, providing profound understanding and more precise forecasts compared to conventional approaches. This thesis assesses the efficacy of four machine learning algorithms, namely Linear Regression, Kernel Ridge Regression, Random Forest Regression, and Support Vector Regression, in assessing patient satisfaction data obtained from the 2021 GP Patient Survey for NHS England. The objective of this study is to determine the primary elements that impact patient satisfaction and evaluate the forecasting skills of different machine learning algorithms. The dataset was carefully prepared and encoded for thorough examination. Linear Regression served as a fundamental tool for gaining understanding of linear associations. Kernel Ridge Regression incorporates non-linear interactions by utilizing kernel functions, which improves adaptability and accuracy. Random Forest Regression is an ensemble method that combines many decision trees to improve robustness and handle complex interactions. Support Vector Regression employs the kernel method to do regression in high-dimensional spaces, effectively capturing intricate patterns. Kernel Ridge Regression had superior performance compared to other models, as it achieved the highest R-squared value and the lowest mean absolute error. This suggests that it is highly effective in capturing both linear and non-linear interactions. An examination of feature importance indicated that appointment scheduling and digital interaction have a considerable impact on patient satisfaction. The simplicity of making appointments and the availability of online booking alternatives were found to be particularly crucial. This study highlights the significance of model interpretability and explainability in healthcare applications. The results indicate that healthcare practitioners should maximize the efficiency of appointment systems and increase digital interfaces to promote patient happiness. Subsequent studies should verify these discoveries using varied datasets and investigate supplementary machine learning approaches. This thesis illustrates the substantial worth of sophisticated analytical methods in augmenting patient contentment.

Date of Award	20 Jul 2024
Original language	American English
Supervisor	Maher Maalouf (Supervisor)